基于稀疏编码的振动信号特征提取算法与实验研究  

Tests and feature extraction algorithm of vibration signals based on sparse coding

作　　者：苗中华[1] 周广兴[1] 刘海宁[2] 刘成良[2] 

机构地区：[1]上海大学机电工程与自动化学院,上海200072 [2]上海交通大学机械与动力工程学院,上海200240

出　　处：《振动与冲击》2014年第15期76-81,118,共7页Journal of Vibration and Shock

基　　金：上海市基础研究重点项目(12JC1404100);上海市科研创新项目(12YZ010);产业化重点项目(11CH-05)资助

摘　　要：针对海量冗余数据中设备状态信息特征提取问题,借鉴生物感知系统"冗余度压缩"的信息处理原则,基于神经科学研究中的稀疏编码算法,提出了连续长时间采样时振动信号有效特征提取方法。介绍了稀疏编码算法及其模型,详细研究了稀疏编码的系数求解和字典学习两大问题。基于人工轴承故障数据集进行了实验研究,实验表明:基于稀疏编码的振动信号特征提取算法不仅能有效提取设备状态特征,而且稀疏特征具有良好的可分性。该方法可用于设备故障诊断,为基于状态的设备智能维护提供有效工具。To solve the problem of status information feature＇s extraction from mass redundant data in a device,an effective feature extraction method of vibration signals long continuously sampled was presented. This method was based on the principle of information processing in a biological perceptual system＇s redundancy compression,and the sparse coding algorithm in neuroscience studying. The two problems of the sparse coding algorithm（ coefficient solving and dictionary learning） were studied in detail. Tests for the proposed method were accomplished based on artificial bearing fault data sets. The results indicated that the proposed vibration signal feature extraction method can not only effectively extract status characteristics from a device＇s information,but also have a good separability from sparse features; this method can be used for an equipment＇s fault diagnosis,and also it provides an effective tool for intelligent maintenance of equipments based on their status.

关 键 词：特征提取 稀疏编码 故障诊断 振动信号分析 

分 类 号：TN911.7[电子电信—通信与信息系统]